minix/kernel/proto.h
Ben Gras 053fa581b5 vm: remove stack handling for signals
. moved to the kernel as the handling was only
	  reading it; the kernel may as well write it too
2012-08-29 17:31:38 +02:00

230 lines
7.7 KiB
C

/* Function prototypes. */
/* FIXME this is a hack how to avoid inclusion conflicts */
#ifdef __kernel__
#ifndef PROTO_H
#define PROTO_H
#include <minix/safecopies.h>
#include <machine/archtypes.h>
#include <a.out.h>
/* Struct declarations. */
struct proc;
struct timer;
/* clock.c */
clock_t get_uptime(void);
void set_timer(struct timer *tp, clock_t t, tmr_func_t f);
void reset_timer(struct timer *tp);
void ser_dump_proc(void);
void cycles_accounting_init(void);
/*
* This functions start and stop accounting for process, kernel or idle cycles.
* It inherently have to account for some kernel cycles for process too,
* therefore it should be called asap after trapping to kernel and as late as
* possible before returning to userspace. These function is architecture
* dependent
*/
void context_stop(struct proc * p);
/* this is a wrapper to make calling it from assembly easier */
void context_stop_idle(void);
int restore_fpu(struct proc *);
void save_fpu(struct proc *);
void save_local_fpu(struct proc *, int retain);
void fpu_sigcontext(struct proc *, struct sigframe *fr, struct
sigcontext *sc);
/* main.c */
#ifndef UNPAGED
#define kmain __k_unpaged_kmain
#endif
void kmain(kinfo_t *cbi);
void prepare_shutdown(int how);
__dead void minix_shutdown(struct timer *tp);
void bsp_finish_booting(void);
/* proc.c */
int do_ipc(reg_t r1, reg_t r2, reg_t r3);
void proc_init(void);
int cancel_async(struct proc *src, struct proc *dst);
int has_pending_notify(struct proc * caller, int src_p);
int has_pending_asend(struct proc * caller, int src_p);
void unset_notify_pending(struct proc * caller, int src_p);
int mini_notify(const struct proc *src, endpoint_t dst);
void enqueue(struct proc *rp);
void dequeue(struct proc *rp);
void switch_to_user(void);
void arch_proc_reset(struct proc *rp);
struct proc * arch_finish_switch_to_user(void);
struct proc *endpoint_lookup(endpoint_t ep);
#if DEBUG_ENABLE_IPC_WARNINGS
int isokendpt_f(const char *file, int line, endpoint_t e, int *p, int
f);
#define isokendpt_d(e, p, f) isokendpt_f(__FILE__, __LINE__, (e), (p), (f))
#else
int isokendpt_f(endpoint_t e, int *p, int f);
#define isokendpt_d(e, p, f) isokendpt_f((e), (p), (f))
#endif
void proc_no_time(struct proc *p);
void reset_proc_accounting(struct proc *p);
void flag_account(struct proc *p, int flag);
int try_deliver_senda(struct proc *caller_ptr, asynmsg_t *table, size_t
size);
/* start.c */
void cstart();
char *env_get(const char *key);
/* system.c */
int get_priv(register struct proc *rc, int proc_type);
void set_sendto_bit(const struct proc *rc, int id);
void unset_sendto_bit(const struct proc *rc, int id);
void fill_sendto_mask(const struct proc *rc, sys_map_t *map);
int send_sig(endpoint_t proc_nr, int sig_nr);
void cause_sig(proc_nr_t proc_nr, int sig_nr);
void sig_delay_done(struct proc *rp);
void kernel_call(message *m_user, struct proc * caller);
void system_init(void);
void clear_endpoint(struct proc *rc);
void clear_ipc_refs(struct proc *rc, int caller_ret);
void kernel_call_resume(struct proc *p);
int sched_proc(struct proc *rp, int priority, int quantum, int cpu);
/* system/do_vtimer.c */
void vtimer_check(struct proc *rp);
/* interrupt.c */
void put_irq_handler(irq_hook_t *hook, int irq, irq_handler_t handler);
void rm_irq_handler(const irq_hook_t *hook);
void enable_irq(const irq_hook_t *hook);
int disable_irq(const irq_hook_t *hook);
void interrupts_enable(void);
void interrupts_disable(void);
/* debug.c */
int runqueues_ok(void);
#ifndef CONFIG_SMP
#define runqueues_ok_local runqueues_ok
#else
#define runqueues_ok_local() runqueues_ok_cpu(cpuid)
int runqueues_ok_cpu(unsigned cpu);
#endif
char *rtsflagstr(u32_t flags);
char *miscflagstr(u32_t flags);
char *schedulerstr(struct proc *scheduler);
/* prints process information */
void print_proc(struct proc *pp);
/* prints the given process and recursively all processes it depends on */
void print_proc_recursive(struct proc *pp);
#if DEBUG_IPC_HOOK
void hook_ipc_msgrecv(message *msg, struct proc *src, struct proc *dst);
void hook_ipc_msgsend(message *msg, struct proc *src, struct proc *dst);
void hook_ipc_msgkcall(message *msg, struct proc *proc);
void hook_ipc_msgkresult(message *msg, struct proc *proc);
void hook_ipc_clear(struct proc *proc);
#endif
/* system/do_safemap.c */
int map_invoke_vm(struct proc * caller, int req_type, endpoint_t end_d,
vir_bytes off_d, endpoint_t end_s, vir_bytes
off_s, size_t size, int flag);
/* system/do_safecopy.c */
int verify_grant(endpoint_t, endpoint_t, cp_grant_id_t, vir_bytes, int,
vir_bytes, vir_bytes *, endpoint_t *);
/* system/do_sysctl.c */
int do_sysctl(struct proc * caller, message *m);
#if SPROFILE
/* profile.c */
void init_profile_clock(u32_t);
void stop_profile_clock(void);
#endif
/* functions defined in architecture-dependent files. */
void prot_init();
void arch_post_init();
void arch_set_secondary_ipc_return(struct proc *, u32_t val);
phys_bytes phys_copy(phys_bytes source, phys_bytes dest, phys_bytes
count);
void phys_copy_fault(void);
void phys_copy_fault_in_kernel(void);
void memset_fault(void);
void memset_fault_in_kernel(void);
#define virtual_copy(src, dst, bytes) \
virtual_copy_f(NULL, src, dst, bytes, 0)
#define virtual_copy_vmcheck(caller, src, dst, bytes) \
virtual_copy_f(caller, src, dst, bytes, 1)
int virtual_copy_f(struct proc * caller, struct vir_addr *src, struct
vir_addr *dst, vir_bytes bytes, int vmcheck);
int data_copy(endpoint_t from, vir_bytes from_addr, endpoint_t to,
vir_bytes to_addr, size_t bytes);
int data_copy_vmcheck(struct proc *, endpoint_t from, vir_bytes
from_addr, endpoint_t to, vir_bytes to_addr, size_t bytes);
phys_bytes umap_virtual(struct proc* rp, int seg, vir_bytes vir_addr,
vir_bytes bytes);
phys_bytes seg2phys(u16_t);
int vm_memset(endpoint_t who,
phys_bytes source, u8_t pattern, phys_bytes count);
int intr_init(int);
void halt_cpu(void);
void arch_init(void);
void arch_boot_proc(struct boot_image *b, struct proc *p);
void cpu_identify(void);
/* arch dependent FPU initialization per CPU */
void fpu_init(void);
/* returns true if pfu is present and initialized */
int is_fpu(void);
void ser_putc(char);
__dead void arch_shutdown(int);
void restore_user_context(struct proc * p);
void read_tsc(u32_t *high, u32_t *low);
int arch_init_profile_clock(u32_t freq);
void arch_stop_profile_clock(void);
void arch_ack_profile_clock(void);
void do_ser_debug(void);
int arch_get_params(char *parm, int max);
void memory_init(void);
void mem_clear_mapcache(void);
void arch_proc_init(struct proc *pr, u32_t, u32_t, char *);
int arch_do_vmctl(message *m_ptr, struct proc *p);
int vm_contiguous(const struct proc *targetproc, vir_bytes vir_buf,
size_t count);
void proc_stacktrace(struct proc *proc);
int vm_lookup(const struct proc *proc, vir_bytes virtual, phys_bytes
*result, u32_t *ptent);
size_t vm_lookup_range(const struct proc *proc,
vir_bytes vir_addr, phys_bytes *phys_addr, size_t bytes);
void delivermsg(struct proc *target);
void arch_do_syscall(struct proc *proc);
int arch_phys_map(int index, phys_bytes *addr, phys_bytes *len, int
*flags);
int arch_phys_map_reply(int index, vir_bytes addr);
reg_t arch_get_sp(struct proc *p);
int arch_enable_paging(struct proc * caller);
int vm_check_range(struct proc *caller,
struct proc *target, vir_bytes vir_addr, size_t bytes);
int copy_msg_from_user(message * user_mbuf, message * dst);
int copy_msg_to_user(message * src, message * user_mbuf);
void switch_address_space(struct proc * p);
void release_address_space(struct proc *pr);
void enable_fpu_exception(void);
void disable_fpu_exception(void);
void release_fpu(struct proc * p);
void arch_pause(void);
short cpu_load(void);
void busy_delay_ms(int ms);
/* utility.c */
void cpu_print_freq(unsigned cpu);
#endif /* __kernel__ */
#endif /* PROTO_H */